ReactEmbed: A Cross-Domain Framework for Protein-Molecule Representation Learning via Biochemical Reaction Networks

2025-01-30

AI Summary

Existing methods inadequately capture joint protein–ligand representations and fail to model their complex biochemical relationships. Method: We propose the first cross-domain unified embedding framework that explicitly leverages biochemical reaction networks as supervisory signals. It integrates pre-trained protein and molecular representations and employs contrastive learning to align reaction pairs, jointly optimizing dual-modality embeddings in a shared latent space. Contribution/Results: Our approach innovatively structures biochemical reaction relationships as a supervised prior—enabling zero-shot generalization (e.g., blood–brain barrier permeability prediction). It achieves state-of-the-art performance across diverse tasks, including drug–target interaction prediction, protein–protein interaction inference, and molecular/protein property prediction. Crucially, experiments demonstrate successful zero-shot transfer to predicting lipid nanoparticle delivery efficiency—a challenging real-world application—validating strong generalizability and practical utility.

Technology Category

• Machine Learning: Representation Learning
• Search and Optimization: Learning to Search
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

The challenge in computational biology and drug discovery lies in creating comprehensive representations of proteins and molecules that capture their intrinsic properties and interactions. Traditional methods often focus on unimodal data, such as protein sequences or molecular structures, limiting their ability to capture complex biochemical relationships. This work enhances these representations by integrating biochemical reactions encompassing interactions between molecules and proteins. By leveraging reaction data alongside pre-trained embeddings from state-of-the-art protein and molecule models, we develop ReactEmbed, a novel method that creates a unified embedding space through contrastive learning. We evaluate ReactEmbed across diverse tasks, including drug-target interaction, protein-protein interaction, protein property prediction, and molecular property prediction, consistently surpassing all current state-of-the-art models. Notably, we showcase ReactEmbed's practical utility through successful implementation in lipid nanoparticle-based drug delivery, enabling zero-shot prediction of blood-brain barrier permeability for protein-nanoparticle complexes. The code and comprehensive database of reaction pairs are available for open use at href{https://github.com/amitaysicherman/ReactEmbed}{GitHub}.

Problem

• Protein-small molecule interactions
• Biochemistry
• Drug discovery

Innovation

• ReactEmbed
• Biochemistry Integration
• Predictive Accuracy